My Authors
Read all threads
I think A-level algorithm by design makes it impossible for any individual student to get A* grade if attend institution where predecessor students 2017-19 did not score A*. Being best student in subject for 4 years ruled out by design.Just get average grade of top 2017-19 pupils
Teachers get to say this student is the top performer in his class. But it is impossible for Ofqual's model to accept that he might be a stronger performer than the *average* of the 2017-19 top student in his subject in his college.
- If top grades were A* (2017), A (2018), A (2019), he gets an A.
- If top grades A* (2017), A* (2018), A (2019), he gets an A*
- If B, B, B, he gets a B

Ofqual marked ghosts of past students, not this student. Impossible for his own attainment to surpass the average ghost
Real example.

Any student being first Oxbridge offer from school for 4 years becomes quite unlikely to get the grades for the place (They depend on past students). If school had many historic Oxbridge offers, the top student will always get in.

Even if using this model [a statistical projection, based on historic students, not current pupils], this deliberate feature of the model (you just can't be best student for 4 years, nor even better than last 3 year average top student) fixable if pre-result appeals from schools
The model has prioritised giving out *the right number* of grades - in aggregate. It is a model whose priority giving out the right number of grades [='credibility'], but not necessarily to the right individual students

Ofqual's model suggests it will get a third of grades wrong
The Ofqual model - by design - does exactly the opposite of what all of the Oxbridge "access" work is supposed to be trying to do. It will *always* - by design - prefer the top private school student to an applicant from the school that doesn't have a history of Oxbridge entry
I hope there will be urgent pressure - from within & outside - to address how the system is deliberately biased against widening access, by systematically marking down [rejecting] students who would widened access. Would like to hear how @arusbridger & his peers will respond
Ofqual have built a model to give what they think are the right average numbers nationally - but their model suggests they get one-third of grades wrong in process, in both directions, as focusing is matching aggregate curve for subjects & institutions
While it is impossible for the top student to out-perform the ghosts, this affects results across the ability range

A school which made year-on-year improvements 2017, 2018, 2019 will now get worse 2020 than 2019 results

It is sometimes argued that "access" approaches to meritocracy can focus too strongly on the exceptional student from poor backgrounds

Ofqual and the govt (through inattention?) have designed a 2020 model that appears designed to systematically disadvantage such students
How the ghosts (the 2017-19 student exam performance) determines what 2020 grades are possible and impossible for any student from that school/college in that subject
The policy choice was
* Hand out roughly right number of grades
* build a 2020 model to v.precisely replicate average 2017-19 attainment gaps between institutions
* Give individuals grades that reflect past exam performance of student "ghosts" & prefer that to any 2020 evidence
This explainer notes a further adjustment mechanism which seems to generate deep arbitrary unfairness at the bottom mirroring the arbitrary ceilings to prevent "over-achievement" (better than the ghosts) at the top. + arbitrary outcoms in between
There should be a review of why the system also insists on distributing U grades arbitrarily

The Education Secretary does now want to "correct" this. It is a core feature of the model designed to distribute 2020 grades.
Ofqual's main priority is to avoid a rise in grades above about 2%. Except private school grades, where A grades were allowed to rise 5%. (But a different decision taken to not arbitrarily correct that, so aggregate stats were not trumps in that case)
Specific case of a student whose top rank in all 3 subjects gets her AAB: those are the fixed *ceiling* grades Ofqual set for any student in these subjects at her college, based on the 2017-19 "ghost" student exam results

The student needed an A* and had an A* CAG. But anything above a B was deemed impossible for any student in her class, based on the exam performances of other people, 12-36 months ago
Ofqual's rule: its impossible for a 2020 student to meet an Oxbridge offer if their class didn't contain past students, 2017-19, who'd have met that offer

Because: past exam results of other people preferred as evidence, outweighing any bespoke assessment of the 2020 candidate
Ofqual's rule (2): impossible for a 2020 student in top 3 of their class to miss an Oxbridge offer grade, as long as their class contained 3 past students, 2017-19, whose grades surpassed it

Ofqual has graded the ghosts of past students - and given their grades to 2020 students
Candidates systematically disadvantaged by Ofqual rule (1) - you can only be as good as the past students who sat at your desk - or advantaged by Ofqual rule (2) - you are automatically as good as past students at your desk - are competing for places
Missing some Tweet in this thread? You can try to force a refresh.

Keep Current with Sunder Katwala

Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!